Clonal Epigenetic Memory in Cancer Cells
Figure 1 – graphical abstract made with CANVA
Abstract
Tumor heterogeneity plays a crucial role in cancer progression and therapy resistance; genetic mutations fail to fully explain it. This study shows how cancer cells can inherit stable transcriptional states through epigenetic memory. By single-cell transcriptomic and methylation analyses, the authors identify distinct DNA methylation regimes that drive non-genetic clonal diversification. In particular, stochastic loss of methylation in DNA leads to rare de-repression events contributing to transcriptional variability. These findings suggest that epigenetics has a role in shaping tumor heterogeneity, indicating that future treatments must target these epigenetic drivers to be truly effective.
Review
Tumor heterogeneity is one of the greatest challenges in modern oncology, complicating diagnosis due to the ability of cancer cells to adapt to changing environments, their therapy resistance and metastatic potential. Traditionally, this diversity has been attributed to the accumulation of genetic mutations. However, genetics alone is not able to explain the full spectrum of variability seen in cancer cells. It provides limited flexibility for rapid adaptation and diversification in a dynamic environment. Consequently, attention was given to epigenetics. Epigenetic modifications are well known to regulate how the genome is read without changing DNA sequence and are known to be reversible, dynamic and stably inherited.
In 2020, researchers at the Weizmann Institute of Science investigated the role of epigenetic memory in tumor evolution, addressing a relevant question1: can cancer cells inherit stable transcriptional and epigenetic states that diversify the population independently of genetic mutations? By using a Luria–Delbrück–style lineage tracking approach, the study demonstrates that cancer cells possess a memory of gene expression programs driven by DNA methylation dynamics, providing a non-genetic mechanism for clonal diversity.
Luria–Delbrück
To investigate whether transcriptional and epigenetic states in cancer cells are transient or stably inherited the authors adopted a Luria-Delbrück approach by coupling single cell transcriptome and methylome measurements with a clonal expansion assay. By isolating individual cells from an initially heterogeneous population and expanding them over multiple divisions, this approach relies on clonal variability to reveal rare molecular states present in founder cells. The study included a panel of lung and fibroblast lines, but the demonstration of heritable transcriptional states was primarily established in the colorectal cancer model HCT116.
This framework revealed that part of the observed transcriptional variability could be attributed to cell-cycle–dependent genes, such as MKI67, TOP2A, and CDC20, whose expression varies naturally through S and M phases. Grouping single cells into metacells based on gene expression similarity revealed that these genes exhibit transient patterns. In contrast, two gene modules, EpCAM and VIM, displayed clonal stability, maintaining consistent expression across metacells. These observations indicate that while transient fluctuations largely arise from cell-cycle dynamics, stable expression patterns are associated with distinct and clonally maintained gene programs.
Transcriptional Memory and Phenotypic Plasticity
The study identified two stable and heritable gene modules linked to the Epithelial-Mesenchymal Transition (EMT): EpCAM and VIM2. Studying their expression was crucial for monitoring the EMT process, which allows tumor cells to lose their adhesion, gain motility, and become metastatic.
The study highlighted the maintenance of EpCAM and VIM expression in both single HCT116 cells and clones along with clonal expansion, suggesting that these modules are stable and inherited. HCT116 cells formed a continuous spectrum ranging from highly Epithelial state (EpCAM-high) to highly Mesenchymal state (VIM-high). Whole-exome sequencing confirmed that clones with similar transcriptional identities (EpCAM – VIM expression) did not share common driver mutations, suggesting that the resulting transcriptional memory is non-genetic. Furthermore, longitudinal tracking at single cell level revealed a slow drift in phenotypes, highlighting epigenetic plasticity, which is stable enough to be inherited across many cell divisions but not permanent like a genetic mutation. This provides a balance between phenotypic stability and the adaptive flexibility essential for tumor evolution.
Dual Regimes of DNA Methylation Dynamics
By a single cell analysis of global DNA methylation profiles of HCT116, the authors validated the presence of two distinct methylation regimes that drive epigenetic diversity: Low-CpG content sites and the High-CpG content sites. The LCG regions, which constitute most of the genome and usually are highly methylated, show a stochastic loss of methylation correlated with late DNA replication timing. This suggests that LCG variability arises from replication-linked passive epimutations, where certain loci progressively undergo hypomethylation, potentially leading to stochastic gene de-repression.
In contrast, the HCG regions, predominantly found in promoters and enhancers, show dynamic methylation instability that is independent of replication timing. Notably, HCG methylation levels correlate with gene expression, particularly in genes involved in cell-cycle regulation suggesting a possible link between proliferation-associated processes and methylation maintenance.
To explain these patterns, the study proposes two kinetic models: “cold” loci, which transmit methylation patterns preserving clonal memory, and “hot” loci, characterized by high turnover and continuous epigenetic drift. These patterns of methylation highlight how cancer cells acquire epigenetic diversity increasing their heterogeneity.
GEMINIs
The researchers developed a screen to identify genes that are strongly repressed in at least 90% of the cell clones but exhibit high expression in only one or a few other clones. These are called GEMINIs (Genes that Escaped Mitotically Inherited Inhibition). GEMINIs include functionally diverse genes, such as transcription factors, long non-coding RNAs, and transmembrane proteins. Notably, five GEMINIs belong to the Cancer-Testis Antigens (CTAs), a group of germline-restricted genes frequently re-expressed in various cancers.
Profiling of DNMT1 and DNMT3B double-knockout (DKO) cells revealed that 40 out of 98 GEMINIs became expressed in most of the cells, confirming that DNA methylation is actively required to maintain their silenced state. The activation is driven by a stochastic, mono-allelic loss of methylation in cis at LCG promoters. Since LCG methylation is generally stable, once this loss occurs the active state is transmitted to daughter cells.
Critically, the clinical relevance of this stochastic loss of methylation was validated through the analysis of bulk RNA-seq data from nearly 400 TCGA colorectal adenocarcinoma tumors, showing that rare de-repression events are more frequently observed in genes identified as GEMINIs. In essence, GEMINIs drive epigenetic clonal diversification, which contributes fundamentally to tumor evolution.
Topological Associating Domains (TADs)
Further, the study revealed that clonally stable gene expression can be shaped by the three-dimensional organization of the genome, particularly through topologically associating domains (TADs). The authors identified 149 genes across 89 TADs, including the embryonic globin cluster (HBE1/HBG) and keratin-associated proteins (KRTAP), that exhibit coordinated clonal co-expression.
Importantly, while TAD de-repression makes multiple genes within a domain more permissive for activation, most cells still exhibit low transcriptional output, indicating that de-repression alone is often not sufficient for strong gene expression. In several cases, high transcriptional output is correlated with the presence of additional trans-acting factors, suggesting that a permissive TAD state must be complemented by further regulatory inputs. Together, these findings support a model in which TADs can exist either in a completely repressed state or an active clonal state, with the latter facilitating gene de–repression but not necessarily a full activation. This mechanism can therefore generate diverse clonal expression patterns among genes that are spatially and functionally linked.
Conclusions
This study fundamentally reshapes our understanding of cancer evolution. It demonstrates that tumors do not only require new genetic mutations to diversify, but also epigenetic variations. Through this analysis it’s shown how cancer cells maintain stable regulatory identities via epigenetic memory. Distinct patterns of methylation at LCG and HCG sites, GEMINIs and TADs highlight the complexity of epigenetic regulation. Together, these observations illustrate how stochastic and spatially coordinated epigenetic events can shape diverse clonal behaviors in genetically identical cells, impacting their metastatic potential, plasticity, and therapy resistance. Nonetheless, most of the findings were derived from ex vivo cell-line systems, and their physiological relevance in vivo remains uncertain. The study also lacks high-resolution single-cell epigenomic analyses in patient tumors, and the functional consequences of these epigenetic states on tumor progression have not been fully tested. Additionally, relying on a certain cell line makes it more challenging to generalize these findings to other cancer types. Future studies could opt for multi-omic single-cell technologies and in vivo models to validate these mechanisms and to assess their prevalence in human tumors in order to determine how clonally stable epigenetic states influence tumor behavior and therapy response. These studies could help to develop epigenetic biomarkers and treatment strategies targeting tumor heterogeneity.
References
- Meir Z, Mukamel Z, Chomsky E, Lifshitz A, Tanay A. Single-cell analysis of clonal maintenance of transcriptional and epigenetic states in cancer cells. Nat Genet. 2020 Jul;52(7):709-718. doi: 10.1038/s41588-020-0645-y.
- Xie Y, Wang X, Wang W, Pu N, Liu L. Epithelial-mesenchymal transition orchestrates tumor microenvironment: current perceptions and challenges. J Transl Med. 2025 Apr 2;23(1):386. doi: 10.1186/s12967-025-06422-5.